A hand-drawing-based government affair unmanned aerial vehicle route generation and multi-scene task adaptation system

Abstract

The application discloses a kind of based on hand-drawing government unmanned aerial vehicle route generation and multi-scene task adaptation system, belong to unmanned aerial vehicle automatic control technical field, it includes intention analysis module, semantic perception module, route rehearsal module, task encapsulation module, for obtaining drawing handwriting data and government task type data and carries out analysis, generates intention guide information, identifies and semantic enhancement perception to current drawing area, generates dynamic semantic potential field and carries out real-time route simulation, generates candidate route segment and carries out virtual space-time resource pre-occupies and conflict pre-judgment;Optimization encapsulation is carried out to the candidate route segment that passes through conflict pre-judgment, generates final route task instruction.The application adopts intention analysis, semantic perception, route rehearsal and task encapsulation combined technical means, can realize the intelligent analysis of user hand-drawing intention and the depth perception to complex scene, to generate unmanned aerial vehicle route that is highly adapted to multi-scene government task quickly, safely.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) automatic control technology, and in particular to a system for generating flight routes and adapting to multiple scenarios for government UAVs based on hand-drawn diagrams. Background Technology

[0002] Unmanned aerial vehicles, commonly known as drones, have been widely used in government sectors such as traffic inspection, environmental monitoring, and emergency mapping due to their advantages of high maneuverability, low cost, and flexible deployment. One of the core components of a government drone system is its mission planning and flight control system. This system is responsible for translating the government's mission requirements into specific flight trajectories and maneuver commands that the drones can execute, and is crucial for achieving automated drone operations.

[0003] Among related technologies, Chinese invention patent CN120088679A discloses a UAV urban management scene patrol and inspection system and method based on geographic information, including: an event processing module, a task management module, a route planning module, a UAV flight module, and a 3D modeling module. It combines geographic information to realize intelligent planning of UAV urban management patrol and inspection tasks. By integrating urban management patrol and inspection business processes, multi-mode task issuance, custom route planning of 3D maps, and task scene management, it achieves a high degree of integration between UAV operation and business management, forming a closed loop of the entire process. This ensures that UAVs can complete tasks in the best state under different scenarios, greatly enhancing the flexibility and adaptability of task issuance.

[0004] However, among the aforementioned technologies, the current interaction methods are rigid, relying on manual planning, which is cumbersome and slow to respond; environmental perception is limited to the geometric level, lacking semantic recognition of obstacles and adaptation to task scenarios; at the same time, the system lacks a forward spatial coordination mechanism, and relies on passive obstacle avoidance when running multiple machines, which poses a great safety hazard; furthermore, the task adaptability is limited, making it difficult to dynamically respond to environmental changes and refined needs, and the intelligent decision-making ability is insufficient. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a hand-drawn-based system for generating flight paths for government drones and adapting them to various scenarios. Employing a combination of intent parsing, semantic awareness, flight path pre-simulation, and task encapsulation, the system enables intelligent parsing of user hand-drawn intents and deep perception of complex scenarios, thereby quickly and securely generating drone flight paths highly adapted to various government tasks.

[0006] The above objectives can be achieved through the following approach: A system for generating flight routes and adapting to multiple scenarios for government drones based on hand-drawn sketches includes: an intent parsing module, used to acquire user-drawn sketch data on an electronic map, the current drawing area, and selected government task type data; performing real-time intent parsing on the sketch data and the government task type data to generate intent guidance information and display it on the electronic map; a semantic perception module, used to respond to the intent guidance information, perform feature recognition and semantic enhancement perception on the current drawing area, and dynamically assign semantic attributes to the identified features in conjunction with the government task type data, generating a dynamic semantic potential energy field based on the semantic attributes of all features; a flight route pre-simulation module, used to perform real-time flight route simulation based on the dynamic semantic potential energy field, generate candidate flight route segments, and submit the candidate flight route segments to the collaborative management center for virtual spatiotemporal resource pre-occupancy and conflict prediction; and a task encapsulation module, used to optimize and encapsulate the candidate flight route segments that have passed conflict prediction after receiving a user confirmation instruction, generating a final flight route task instruction and issuing it for execution.

[0007] Optionally, the intent parsing module includes: an intent inference unit, used to parse the morphological features of the drawn handwriting data and, in conjunction with the government task type data, infer the user's potential operational intent; a rule loading unit, used to load associated flight rules and operational constraints from a preset scenario strategy knowledge base according to the potential operational intent; and a guidance generation unit, used to generate and render intent guidance information in real time based on the flight rules and operational constraints, as well as the identified key ground feature information related to the current operational intent.

[0008] Optionally, loading associated flight rules and operational constraints from a preset scenario strategy knowledge base includes: performing multi-level semantic retrieval and context matching in the scenario strategy knowledge base based on the potential operational intent and the geographical environmental features of the currently drawn area, and dynamically constructing an operational scenario context graph; extracting flight behavior templates that match the government task type based on the operational scenario context graph; obtaining real-time perceived environmental state information, and performing online correction and elastic scaling of behavior parameters on the flight behavior templates to form flight rules and operational constraints that take effect in real time.

[0009] Optionally, the semantic perception module includes: a topology construction unit, used to identify basic geographic elements within the currently drawn area, and combine the intent guidance information to perform enhanced perception and structured analysis to construct a topology network reflecting the relationship logic between geographic features; a semantic matching unit, used to dynamically match semantic attributes of the geographic nodes identified in the topology network and adapt them to the task scenario based on the government task type data, forming a semantic geographic feature set; and a potential energy field generation unit, used to calculate the semantic force field distribution and generate a dynamic semantic potential energy field based on the semantic role association degree of each element in the semantic geographic feature set and its influence weight on task execution.

[0010] Optionally, the route pre-simulation module includes: a route exploration unit, used to explore routes based on potential energy gradients in the dynamic semantic potential energy field and generate an initial path point sequence; a route calibration unit, used to perform smoothing optimization and feasibility verification based on the initial path point sequence to form a basic route segment; and an evaluation and submission unit, used to calculate the performance indicators of the basic route segment, assign corresponding priority weights to the performance indicators according to the government task type data, obtain a weighted evaluation result for configuration selection, generate candidate route segments, and submit them to the collaborative management center for virtual spatiotemporal resource pre-occupancy and conflict prediction.

[0011] Optionally, submitting the virtual spatiotemporal resource pre-occupancy and conflict prediction to the collaborative management center includes: extracting spatiotemporal resource occupancy information contained in the candidate flight route segment, wherein the spatiotemporal resource occupancy information includes the planned occupancy time window and the three-dimensional contour of the flight airspace; sending the spatiotemporal resource occupancy information to the collaborative management center, calculating the spatiotemporal overlap with the spatiotemporal resource occupancy information of other tasks, and obtaining the spatiotemporal overlap; if the spatiotemporal overlap exceeds a preset safety threshold, generating conflict prediction alarm information, and returning the conflict prediction alarm information as feedback information to adjust the intent guidance information.

[0012] Optionally, adjusting the intent guidance information includes: generating a visual spatiotemporal conflict heatmap and dynamic adjustment suggestion information based on the spatiotemporal overlap area, conflict task type, and current task scenario contained in the conflict prediction alarm information; overlaying and rendering the visual spatiotemporal conflict heatmap onto the currently drawn area of ​​the electronic map, and reconstructing the intent guidance information in real time based on the dynamic adjustment suggestion information to obtain the reconstructed intent guidance information; responding to the user's secondary confirmation or modification operation on the reconstructed intent guidance information, triggering a new round of dynamic semantic potential energy field calculation and candidate route segment generation, until the spatiotemporal overlap is lower than the safety threshold.

[0013] Optionally, the task encapsulation module includes: a conflict determination unit, used to receive the conflict prediction result of the candidate route segment returned by the collaborative management center, and trigger the optimization process if the conflict prediction result is passed; a route fine-tuning unit, used to fine-tune and optimize the waypoint order and flight parameters in the candidate route segment based on the optimization process, and generate the optimized final route; and a protocol encapsulation unit, used to encapsulate the waypoint coordinates and task instructions of the optimized final route according to the protocol format that the target UAV can parse, and generate the final route task instructions.

[0014] Optionally, the fine-tuning and optimization of the waypoint order and flight parameters in the candidate route segment includes: adjusting the waypoint density based on the dynamic semantic potential energy field, merging redundant waypoints and enhancing key area coverage to form a semantically coherent waypoint sequence; adjusting the flight altitude and speed parameters between waypoints according to the semantically coherent waypoint sequence and in conjunction with the environmental state information to achieve trajectory smoothing and energy consumption optimization; and optimizing the waypoint execution order in terms of timing and airspace alignment based on the fine-tuned flight parameters and multi-aircraft collaborative constraints to generate the optimized final route.

[0015] Based on the same inventive concept, this invention also provides a method for generating flight routes and adapting to multiple scenarios for government drones based on hand-drawn sketches. The method includes: acquiring user-generated handwriting data on an electronic map, the current drawing area, and selected government task type data; performing real-time intent analysis on the handwriting data and the government task type data to generate intent guidance information and display it on the electronic map; responding to the intent guidance information, performing feature recognition and semantic enhancement perception on the current drawing area, and dynamically assigning semantic attributes to the identified features in conjunction with the government task type data, generating a dynamic semantic potential energy field based on the semantic attributes of all features; performing real-time flight route simulation based on the dynamic semantic potential energy field to generate candidate flight route segments, and submitting the candidate flight route segments to the collaborative management center for virtual spatiotemporal resource pre-occupancy and conflict prediction; upon receiving a user confirmation instruction, optimizing and encapsulating the candidate flight route segments that have passed the conflict prediction, generating a final flight route task instruction, and issuing it for execution.

[0016] Compared with the prior art, the present invention has the following advantages: 1. This invention uses a hand-drawn interactive method combined with an intent parsing module to directly transform the user's intuitive drawing operations into structured task intents, simplifying the drone mission assignment process, reducing the professional skill requirements for operators, enabling frontline government personnel to quickly deploy drone missions, and improving the planning efficiency of emergency response and daily operations.

[0017] 2. This invention introduces the concept of dynamic semantic potential energy field. Through the semantic perception module, it deeply understands the working environment, not only identifying the physical attributes of ground objects, but also assigning semantic attributes to them in combination with the type of government affairs task. This enables the route generation process to proactively seek advantages and avoid disadvantages, and the route itself is more in line with the actual needs of the task and the inherent logic of the environment. Thus, it realizes the upgrade of route planning from pure geometric path optimization to intelligent and context-aware decision control, improving flight safety and the accuracy of task execution.

[0018] 3. This invention designs an interactive mechanism between route rehearsal and collaborative management center. By pre-occupying virtual spatiotemporal resources and predicting conflicts, the airspace conflict risks of multi-UAV collaborative operations are pre-investigated and resolved before mission execution. This transforms passive real-time obstacle avoidance into proactive planned avoidance, providing safety assurance for large-scale, high-density government UAV cluster collaborative operations and effectively improving the utilization efficiency of airspace resources and the reliability of overall operations.

[0019] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a framework diagram of a hand-drawn government drone route generation and multi-scenario task adaptation system according to an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram of the structure of a hand-drawn government drone route generation and multi-scenario task adaptation system according to an embodiment of the present invention.

[0023] Figure 3 This is a three-dimensional schematic diagram of the dynamic semantic potential energy field according to an embodiment of the present invention.

[0024] Figure 4 This is a potential energy field-based route exploration path diagram according to an embodiment of the present invention.

[0025] Figure 5 This is a flowchart illustrating a method for generating flight routes and adapting to multiple scenarios for government drones based on hand-drawn diagrams, according to an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Reference Figure 1 One embodiment of the present invention proposes a system for generating and adapting government drone flight paths based on hand-drawn sketches. It employs a combination of intent parsing, semantic perception, flight path pre-simulation, and task encapsulation to achieve intelligent parsing of user hand-drawn intentions and deep perception of complex scenarios, thereby quickly and safely generating drone flight paths highly adapted to various government tasks.

[0028] The system described in this embodiment specifically includes: The intent parsing module is used to obtain the user's handwriting data, the current drawing area, and the selected government task type data on the electronic map, perform real-time intent parsing on the handwriting data and the government task type data, generate intent guidance information, and display it on the electronic map. Optionally, the intent parsing module includes: The intent inference unit is used to analyze the morphological features of the handwriting data and, in combination with the government task type data, infer the user's potential work intent. The rule loading unit is used to load associated flight rules and operational constraints from a preset scenario strategy knowledge base according to the potential operational intent. The guidance generation unit is used to generate and render intent guidance information in real time based on the flight rules and operational constraints, as well as the identified key ground features related to the current operational intent.

[0029] Specifically, such as Figure 2As shown, the process begins with the intent inference unit, which receives handwriting data collected from the front end. This data is a sequence of coordinate points including timestamps, with a sampling rate typically set between 50-100 Hz. First, the original coordinate sequence undergoes noise reduction and smoothing, using a moving average filter to filter continuously collected handwriting coordinate points in real time, eliminating high-frequency noise introduced by hand tremors or device sampling errors. Then, a key point extraction and shape simplification algorithm, such as the Douglas-Puk algorithm, is used. This involves recursively obtaining the vertical distance from each point to the lines connecting the start and end points, retaining key coordinate points whose distance exceeds a set threshold (e.g., screen pixel distance or actual geographic distance of 5-15 pixels), and discarding redundant intermediate points. Based on this, a classifier combining template matching and a lightweight convolutional neural network is run to identify the morphological features of the handwriting, such as straight lines, closed polygons, circles, or open curves. Simultaneously, data for the user-selected government task type, such as "farmland inspection" or "traffic violation capture," is obtained. A pre-defined decision matrix is ​​used to correlate identified morphological features with government task type data to infer the user's potential operational intent. This matrix uses handwriting morphological features (such as straight lines, closed polygons, and circles) and government task type data (such as farmland inspection and illegal construction inspection) as indices for input dimensions. Each cell in the matrix corresponds to one or more potential operational intents, such as aerial photography covering a designated area. For example, when the government task type data is "regional surveying" and the morphological feature is "closed polygon," the potential operational intent is determined to be "aerial photography covering a designated area." To ensure the accuracy of the inference, an intent matching confidence score is calculated. The calculation of the intent matching confidence score... ,have: ; in, The score representing the matching score between handwriting pattern and standard operating procedure is output by the classifier and ranges from 0 to 1. The priority score for the current government task type is determined by system configuration; It is the score of the fit between the environmental characteristics of the current mapping area and the intention of the operation. For example, the fit is low when carrying out large-area mapping in a densely built area. , , The weighting coefficients for each score are set to a total of 1, and their values ​​are pre-calibrated based on historical task data and expert experience. Only when C exceeds a preset threshold, such as 0.8, is the potential operational intent confirmed. Next, the rule loading unit is triggered, using the confirmed potential operational intent as a query index to retrieve the associated rule set from a preset scenario strategy knowledge base. This knowledge base is constructed using a graph database, where nodes represent rules or constraints, and edges represent their dependencies and triggering relationships. The loading process is not simply reading; it combines the geographical features of the currently drawn area with real-time weather information to parameterize and instantiate templated flight behaviors. For example, for the intent of "specified area coverage aerial photography," the unit loads a "bow" or "return" shaped flight pattern and automatically calculates the flight path spacing based on the area and shape, which must meet a minimum image overlap requirement of 70%. Simultaneously, operational constraints such as minimum safe flight altitude, maximum flight speed, and camera gimbal pitch angle range (e.g., -90 degrees to -30 degrees) are loaded. Finally, the guidance generation unit is activated, receiving flight rules and operational constraints output by the rule loading unit, as well as key geographical information related to the current operational intent obtained from the semantic perception module, such as high-voltage power line towers and no-fly zone boundaries. It performs real-time rendering on the currently drawn area of ​​the electronic map, generating intent guidance information. For example, based on minimum safe flight altitude constraints and tall building geographical information, it renders a virtual, semi-transparent safe altitude surface; based on the loaded flight paradigm and flight path spacing, it dynamically generates a recommended flight path skeleton within the polygon drawn by the user. This intent guidance information is updated in real-time as the user continues to draw or modify their handwriting until the user confirms the input, thus completing the entire intent parsing process.

[0030] For example, a law enforcement officer selects the "illegal building inspection" task as the administrative task type data and circles a target building on the electronic map. The intent parsing module is then activated. Its intent inference unit analyzes the handwriting data, identifies the shape as a "closed polygon," and, combined with the administrative task type data, infers the user's potential operational intent as "to conduct multi-angle surround photography of the target area." To ensure accuracy, the intent matching confidence score is calculated. In this verification example, the handwriting shape matches the standard operating mode of surround photography very well, with a matching score of 0.9 output by the classifier; the current "illegal building inspection" task is set as a high priority, with a priority score of 0.9; the currently drawn area is an urban building area, which has a high degree of fit for conducting surround photography, with a fit score of 0.95. Weights , , Given values ​​of 0.5, 0.2, and 0.3 respectively, the calculated values ​​are... The value exceeds a preset threshold of 0.8, confirming the intent. Next, the rule loading unit loads the "around point of interest" flight rule from the scene strategy knowledge base based on this intent. Finally, the guidance generation unit renders a circular dashed circle around the target building on the map as intent guidance information to assist the user's planning. By performing in-depth intent analysis on simple, unstructured hand-drawn input from users, complex and professional work patterns can be automatically inferred, lowering the professional threshold for operators and improving the response speed and planning efficiency of government tasks.

[0031] Optionally, loading the associated flight rules and operational constraints from the preset scenario strategy knowledge base includes: Based on the potential task intent and the geographical features of the currently drawn area, multi-level semantic retrieval and context matching are performed in the scene strategy knowledge base to dynamically construct a task scene context graph; Based on the operational scenario context graph, flight behavior templates that match the administrative task type are extracted; The system acquires real-time environmental status information and performs online corrections and flexible scaling of behavior parameters on the flight behavior template to form real-time effective flight rules and operational constraints.

[0032] Specifically, this is first accomplished by dynamically constructing a task scenario context graph. The potential task intent output by the intent parsing module and the geographic environmental features of the currently drawn area are used as input for a joint query. Multi-layer semantic retrieval is performed in a scenario strategy knowledge base constructed using graph database technology. The nodes in this knowledge base include feature types, such as buildings and water bodies; task intents, such as inspection and surveying; and behavioral constraints, such as height restrictions and no-entry restrictions. Edges represent the logical relationships between them, such as "contains," "proximity," and "requirement." The retrieval process first locates the root node matching the potential task intent. Then, based on the geographic environmental features of the current area, such as "high-density urban area" or "riverbank," a breadth-first or depth-first subgraph search is performed around the root node. Finally, pruning is performed to form a task scenario context graph containing all relevant entities and their relationships. Subsequently, based on the constructed task scenario context graph, flight behavior templates matching the government task type are extracted. Flight behavior templates are a set of parameterized basic action sequences, such as "grid scan template," "around point of interest template," or "round trip template." Within the graph, based on the administrative task type node, strongly correlated flight behavior template nodes are retrieved along the relational edges. For example, a "traffic violation capture" task in the "urban main road" graph context will preferentially match the "along-route return template" rather than the "grid scanning template." Finally, real-time environmental status information is acquired, and the selected flight behavior templates are corrected online and their behavior parameters are flexibly scaled to form real-time effective flight rules and operational constraints. Real-time environmental status information is obtained through airborne sensors or third-party data interfaces; key parameters include wind speed, wind direction, and the horizontal accuracy factor of satellite positioning signals. This information is used to dynamically adjust key flight parameters in the templates, such as flight speed. The adjusted flight speed is then calculated. ,have: ; in, It is the base flight speed obtained from the flight behavior template; It is the real-time wind speed. The angle between the heading and the wind direction; This refers to the drag coefficient, with a typical value of 0.2-0.5. and These are the minimum and maximum value functions, respectively; H is the horizontal precision factor, typically ranging from 0.5 (excellent) to 10.0 (poor). Through this dynamic correction, a set of flight rules and operational constraints that are fully adapted to the current mission intent, geographical scenario, and real-time environment is finally output for use by subsequent modules.

[0033] For example, based on the intent of "surrounding for detailed shooting," the rule loading unit retrieves information from the scene strategy knowledge base and dynamically constructs a scenario context graph containing nodes such as "tall buildings" and "signal obstruction." Based on this, it extracts a matching "surrounding point of interest" flight behavior template with a base flight speed of 5 meters per second. The key lies in online correction, obtaining a real-time wind speed of 4 meters per second, a heading angle of 0 degrees with the wind direction, a drag coefficient of 0.4, and a GPS horizontal accuracy factor of 1.5. First, the speed adjustment coefficient is calculated. Then, the adjusted flight speed was calculated. (meters per second). Ultimately, a set of flight rules and operational constraints, including adjusted speed and "flight altitude must not be lower than 20 meters from surrounding buildings," are generated in real time and passed to subsequent modules. By dynamically constructing an operational scenario context graph, it is possible to go beyond simple intent matching, deeply understand the complex constraints of the task in a specific environment, and improve flight safety and the success rate of the final mission.

[0034] The semantic perception module is used to respond to the intent guidance information, perform feature recognition and semantic enhancement perception on the currently drawn area, and dynamically assign semantic attributes to the identified features in combination with the government task type data, and generate a dynamic semantic potential energy field based on the semantic attributes of all features. Optionally, the semantic awareness module includes: The topology construction unit is used to identify the basic geographic features within the currently drawn area, and combine the intent guidance information to perform enhanced perception and structured analysis to construct a topology network that reflects the relationship logic between geographic features; The semantic matching unit is used to dynamically match semantic attributes of the various feature nodes identified in the topological network with the task scenario based on the government task type data, forming a semantic feature set; The potential energy field generation unit is used to calculate the semantic force field distribution and generate a dynamic semantic potential energy field based on the semantic role association degree of each element in the semanticized feature set and its influence weight on task execution.

[0035] Specifically, this process begins with a topology building unit. This unit first calls a pre-trained remote sensing image segmentation model, such as a deep learning network based on the U-Net architecture, to perform pixel-level land feature classification on high-resolution satellite or aerial images of the currently mapped area, identifying basic geographic features such as buildings, roads, water bodies, and vegetation. The initial recognition confidence threshold is typically set above 0.9. The input to the deep learning network is a high-resolution RGB or fused multi-band remote sensing image tile of the currently mapped area, and the output is a probability distribution map of land feature categories for each pixel. The training process uses a large-scale labeled remote sensing image dataset. Subsequently, enhanced perception is achieved by combining intent guidance information generated by the intent parsing module. For example, if the intent is "high-voltage line inspection," this unit will increase the sensitivity to the recognition of specific land features such as poles and cables. The identified land features are abstracted as nodes, and through geometric relationships such as adjacency, intersection, and containment between them, an undirected graph-structured topological network is constructed. This network is stored in the form of an adjacency matrix or adjacency list, providing a structured foundation for subsequent semantic assignment. Next, the semantic matching unit is activated. This unit takes government task type data as input and queries a task knowledge base built on ontology. This knowledge base associates different government tasks with the types of features they focus on and the semantic roles these features play in the task. For example, for the "traffic violation capture" task, the "road" node in the topology network will be matched with the semantic attribute of "main work area," while the "high-rise building" node may be assigned the semantic attribute of "signal obstruction risk source." By traversing all nodes in the topology network and performing this matching process, a semantically defined feature set is generated, where each feature not only has its physical location and morphological information but also carries a set of semantic attribute labels adapted to the task scenario. Finally, the potential energy field generation unit is responsible for transforming the discrete semantically defined feature set into a continuous, dynamic semantic potential energy field that guides flight path generation. This unit discretizes the space into a three-dimensional raster map, with a raster resolution typically set to 1 to 5 meters. For each raster point, its potential energy value is calculated by superimposing the effects of all semantically defined features. The total potential energy value at raster point p is... ,have: ; Where N is the total number of semantically coded ground features; The semantic influence weight of the i-th land feature can be set between -1000 and 1000. These are the coordinates of a grid point in three-dimensional space. This represents the geometric center or influence source location of the i-th feature; To calculate the Euclidean distance between two points; The influence radius of the i-th ground feature is determined by its physical size and safety margin. By calculating for all grid points within the region, a three-dimensional potential energy matrix, or dynamic semantic potential energy field, is generated. This field visually represents safe and efficient flight areas (low-potential-energy valleys) and dangerous or irrelevant areas to be avoided (high-potential-energy peaks). For example... Figure 3 As shown, each point in space has a "potential energy value". The level of potential energy represents the suitability of the drone flight. There is a "key avoidance area" that needs to be avoided, such as a school, which forms a towering "potential energy peak", i.e., a high-risk area. At the same time, there is a "core survey target", such as an illegal building, which forms a deep "potential energy depression", i.e., a preferred area.

[0036] For example, the topology building unit identifies target residential buildings, nearby primary schools, and roads through image segmentation and constructs a topological network. The semantic matching unit, based on the "illegal building inspection" task, assigns semantic attributes to features, designating the target rooftop as the "core inspection target" and the primary school as a "key avoidance area." Finally, the potential energy field generation unit transforms these semantics into a quantified dynamic semantic potential energy field. In this verification example, the primary school is assigned a high repulsive weight of 800, with an influence radius of 50 meters; the illegal building is assigned a low attractive weight of -1000, with an influence radius of 30 meters. For a grid point p in space, 40 meters from the primary school and 25 meters from the illegal building, its total potential energy value is... By calculating the entire area, a three-dimensional potential energy matrix was generated, forming "potential energy peaks" (high-risk areas) around primary schools and "potential energy depressions" (preferred areas) around targets. By dynamically assigning semantic attributes to ground features and constructing a potential energy field, the system no longer simply identifies "a building" and "a school," but understands "this is the target to be photographed" and "this is a sensitive area to be avoided." This quantified potential energy field provides clear guidance for subsequent flight path planning, laying a solid foundation for generating safe, efficient, and compliant flight paths.

[0037] The route pre-simulation module is used to perform real-time route simulation based on the dynamic semantic potential energy field, generate candidate route segments, and submit the candidate route segments to the collaborative management center for virtual spatiotemporal resource pre-occupation and conflict prediction. Optionally, the route pre-simulation module includes: The path exploration unit is used to explore paths based on the potential energy gradient in the dynamic semantic potential energy field and generate an initial path point sequence. The route calibration unit is used to perform smoothing optimization and feasibility verification based on the initial path point sequence to form a basic route segment. The evaluation and submission unit is used to calculate the performance indicators of the basic route segment, assign corresponding priority weights to the performance indicators according to the government task type data, obtain weighted evaluation results for configuration selection, generate candidate route segments, and submit them to the collaborative management center for virtual spatiotemporal resource pre-occupancy and conflict prediction.

[0038] Specifically, the path exploration unit uses a dynamic semantic potential energy field—a 3D grid map—as its search space and employs an improved A* algorithm for path exploration. The algorithm starts from the user-specified starting point and searches towards the target point or region. Its cost function is defined as the sum of the actual cost from the starting point to the current grid node n and the estimated cost from node n to the target point. The actual cost includes not only the geometric length of the path but, more importantly, the accumulated potential energy values ​​of all grids traversed along the path, causing the algorithm to instinctively avoid high-potential-energy regions during exploration. The estimated cost uses Euclidean distance in 3D space as a heuristic function. For example, to calculate the estimated total cost from the starting point through the current node n to the target point... ,have: ; in, This refers to the current grid node; This refers to the i-th grid point on the path; The Euclidean distance between adjacent nodes; This is the path length weighting coefficient, typically with a value of 1.0; This is the potential energy accumulation weighting coefficient, typically ranging from 10.0 to 100.0; For heuristic weighting coefficients, when When the value is equal to 1, the algorithm guarantees to find the optimal path. When the value is greater than 1, the algorithm searches faster, but may not find the optimal path. When the value is less than 1, the search is more thorough, but the efficiency is reduced. For the target point; The number of grid cells along the path. Through iterative expansion within eight or twenty-six neighborhoods, the cell ultimately generates an initial sequence of path points consisting of three-dimensional coordinates, distributed along the "valleys" of lowest potential energy in the potential field. For example... Figure 4As shown, the dynamic semantic potential energy field is represented by two-dimensional contour lines, overlaid with an exploration path from the starting point to the ending point. Darker areas represent higher potential energy and greater danger, while lighter areas represent lower potential energy and greater safety. The generated "initial path" clearly bypasses high-potential-energy areas, proceeding along "valleys" of lower potential energy. Subsequently, the flight path calibration unit receives this initial path point sequence. This unit first applies a B-spline interpolation algorithm to fit the initial path point sequence, generating a continuous and high-order differentiable smooth curve, ensuring the continuity of the trajectory's speed and acceleration. Next, a feasibility check is performed, verifying whether the radius of curvature at any point on the smooth curve is greater than the UAV's minimum turning radius (typically 5 to 10 meters for a medium-sized multi-rotor UAV). The unit also checks whether the trajectory's climb and descent gradients are within the UAV's maximum climb / descent rate, such as ±5 meters per second. For segments that do not meet the constraints, the unit performs local replanning by adjusting the B-spline control points until the entire flight path satisfies the UAV's kinematic constraints. This processing creates a basic flight path segment. Finally, the evaluation and submission unit performs a quantitative evaluation and decision-making process on the generated basic flight path segment. This unit calculates multiple performance indicators for the basic flight path segment, primarily including total distance, cumulative potential energy of the route, and mission coverage. Subsequently, based on the government task type data, the corresponding priority weights are retrieved from the configuration library, and the weighted evaluation results are calculated. The weighted evaluation score is then calculated. ,have: ; in, This refers to the total voyage length; Accumulate potential energy for the flight path; For task coverage; , , These are the priority weights for flight distance, flight risk, and mission completion quality, respectively, and their sum is 1. The specific value is determined by the type of government mission data; for example, emergency reconnaissance missions will have higher priority weights. The weight is higher, while security patrol tasks will have a higher weight. Weight; It is a normalized function; This refers to the original value of a certain performance metric. , The minimum and maximum values ​​of this indicator among all candidate routes are respectively selected. The basic route segment with the best score is selected as the candidate route segment, packaged, and containing detailed four-dimensional spatiotemporal information. It is then submitted to the collaborative management center for subsequent virtual spatiotemporal resource pre-occupancy and conflict prediction processes.

[0039] For example, based on the constructed dynamic semantic potential energy field, the route pre-simulation module is launched. The path exploration unit uses an improved A* algorithm to search for paths in the potential energy field. Its cost function integrates the geometric length of the path and the accumulated potential energy value along the way, enabling the algorithm to automatically explore along the "valley" with the lowest potential energy and avoid high-risk areas. The generated initial path point sequence is smoothed by the route calibration unit using B-spline interpolation to ensure flight feasibility. Finally, the evaluation submission unit calculates the route performance. In this verification example, for the evidence collection task, the total flight distance is 550 meters, the cumulative potential energy of the route is -8500, and the task coverage rate, i.e., the coverage of illegal buildings from multiple angles, is 95%; the weight of task coverage rate is set to 0.6, the risk weight is 0.3, and the flight distance weight is 0.1. If there is only this one candidate route, its normalized function N(x) after processing each index value is N(L) = 0.8, N( )=0.2, N( The weighted evaluation score is 0.9. The route with the highest score was selected as a candidate route segment, and its four-dimensional spatiotemporal information was encapsulated and submitted to the collaborative management center for conflict prediction. By combining the A* algorithm with potential energy fields and a weighted evaluation mechanism, the route selection can closely align with the core requirements of government tasks, thereby generating truly "task-aware" candidate routes, providing high-quality input for subsequent collaborative scheduling and final execution.

[0040] Optionally, the step of submitting the virtual spatiotemporal resource pre-occupancy and conflict prediction to the collaborative management center includes: Extract the spatiotemporal resource occupancy information contained in the candidate route segments, the spatiotemporal resource occupancy information including the planned occupancy time window and the three-dimensional contour of the flight airspace; The spatiotemporal resource occupancy information is sent to the collaborative management center, and the spatiotemporal overlap is calculated with the spatiotemporal resource occupancy information of other tasks to obtain the spatiotemporal overlap. If the spatiotemporal overlap exceeds a preset safety threshold, a conflict prediction alarm is generated and returned as feedback information to adjust the intent guidance information.

[0041] Specifically, the process first requires the precise extraction of spatiotemporal resource occupancy information contained in the candidate flight path segments. This involves parsing the coordinates of all waypoints and their associated timestamps within the candidate flight path segments to determine the planned occupancy time window—the time interval from the estimated departure time of the first waypoint to the estimated arrival time of the last waypoint—with an additional 5 to 10 seconds of safety redundancy. Simultaneously, a three-dimensional airspace contour is constructed around the entire flight path. This contour is typically modeled as a tubular space composed of continuous cylinders or capsules, with its radius determined based on the UAV's size, positioning accuracy error (usually 3 to 5 meters), and flight speed, ensuring that the UAV's actual flight trajectory has a greater than 99.9% probability of being contained within this contour. Subsequently, the encapsulated spatiotemporal resource occupancy information, including the task identifier, planned occupancy time window, and geometric description of the three-dimensional airspace contour, is sent to the collaborative management center via a standardized airspace management protocol. Upon receiving the request, the center compares it with the spatiotemporal resource occupancy information of all other pre-occupied or currently executing tasks in the database. For any pair of tasks, the first step is to determine the overlap of time windows. If the time windows do not intersect, no conflict is considered. If the time windows intersect, a geometric intersection operation is performed on the three-dimensional contours of the spatial domain to calculate the volume of the overlapping region. This geometric intersection operation can be achieved by calculating the Boolean intersection volume of two sets of capsules or cylinders. This can be done by discretizing the airspace occupied by the two flight paths into three-dimensional voxel sets using a spatial voxelization method; the intersection volume is then the volume of the voxel intersection. Finally, a quantified spatiotemporal overlap degree is generated based on the calculation results, and conflict prediction is made accordingly. The calculation of the spatiotemporal overlap degree... ,have: ; in, and These are the total volumes of the three-dimensional spatial contours of the two tasks, respectively. It is the intersection volume of the three-dimensional contours of the two task spatial domains; It is the intersection duration of the time windows occupied by the two task plans; and These represent the total planned time windows for each of the two tasks. The formula integrates the degree of spatial and temporal overlap. If the calculated spatiotemporal overlap exceeds a dynamically set safety threshold based on airspace security levels—for example, this threshold might be as low as 0.01 in densely populated urban areas, but relaxed to 0.1 in open areas—a conflict prediction alarm is generated. This alarm not only includes the conflict determination result but also the specific time period, spatial area, and task information of the opposing party in the conflict. This structured feedback is returned to the system to trigger subsequent intent guidance information adjustment processes, thus forming a closed-loop planning and avoidance mechanism.

[0042] For example, candidate route segments are submitted to the collaborative management center for conflict prediction. First, the spatiotemporal resource occupancy information of the route segment is extracted. According to the route planning, the expected occupancy time window is from 10:05:15 AM to 10:08:57 AM, with a total duration of... The time was 222 seconds. Simultaneously, a three-dimensional contour of a pipe-shaped airspace with a radius of 4 meters was constructed around the flight path, and its total volume was calculated. The total space and time resource occupancy is 3140 cubic meters. This information was packaged and sent to the collaborative management center. Coincidentally, another drone was carrying out a "regional air quality monitoring" mission initiated by the environmental protection department, and its pre-occupied space and time resource information was also already in the central database. The planned time window for this mission was from 10:07:30 to 10:10:25, with a total duration of... The duration is 175 seconds; the total volume of its three-dimensional spatial contour. The volume is 4825 cubic meters. After receiving the request for the "illegal building inspection" task, the collaborative management center immediately calculates the spatiotemporal overlap. First, it determines the overlap of time windows. The time windows of task A [10:05:15, 10:08:57] and task B [10:07:30, 10:10:25] intersect, with the intersection being [10:07:30, 10:08:57]. The intersection duration is... The time was 87 seconds. Next, a geometric intersection calculation of the three-dimensional airspace contours was performed. The results showed that the two flight paths highly overlapped above a certain intersection, with an intersection volume of [missing information]. The volume is 162.5 cubic meters. Finally, the quantified spatiotemporal overlap is calculated. Although these values ​​overlap, they do not exceed the preset safety threshold of 0.05. Therefore, the collaborative management center determines it as "low risk, pass," and feeds this prediction back to the planning system for the "illegal building inspection" task. In this example, we assume another scenario: if the intersecting volume is 355 cubic meters, the calculated result... If the threshold exceeds 0.05, a conflict is identified, and a conflict prediction alarm is generated, returning feedback information including the conflict time period, spatial area, and the other party's task type (i.e., air quality monitoring). By reviewing flight path planning from a macro perspective of multi-task collaboration, crucial data is provided for subsequent flight path fine-tuning or user decisions, ensuring airspace safety and operational order in multi-UAV collaborative operation environments.

[0043] Optionally, adjusting the intent guidance information includes: Based on the spatiotemporal overlap area, conflict task type, and current task scenario contained in the conflict prediction and alarm information, a visual spatiotemporal conflict heat map and dynamic adjustment suggestion information are generated. The visualized spatiotemporal conflict heatmap is overlaid and rendered onto the currently drawn area of ​​the electronic map, and the intent guidance information is reconstructed in real time based on the dynamic adjustment suggestion information to obtain the reconstructed intent guidance information. In response to the user's secondary confirmation or modification of the reconstructed intent guidance information, a new round of dynamic semantic potential energy field calculation and candidate route fragment generation is triggered until the spatiotemporal overlap is lower than the safety threshold.

[0044] Specifically, the process begins upon receiving a conflict prediction alarm from the collaborative management center. First, the alarm is analyzed, extracting the geometric description of the spatiotemporal overlap area, the conflict task type, and the conflict severity level. Based on this data, a visual spatiotemporal conflict heatmap is generated. Simultaneously, based on the built-in conflict resolution strategy library and the current task scenario, dynamic adjustment suggestions are generated. For example, if the conflict type is highly overlapping, the suggestion might be "Adjust the flight altitude to above XX meters"; if it's a time conflict, it might be "Delay the task start time by XX minutes." Then, the generated spatiotemporal conflict heatmap is overlaid and rendered onto the current drawing area of ​​the electronic map, and the dynamic adjustment suggestions are presented to the user in the form of text boxes or voice announcements. The heatmap overlay rendering uses transparency blending technology to ensure that it clearly indicates the conflict risk without obscuring the underlying map information. At the same time, based on the dynamic adjustment suggestions, the original intent guidance information is reconstructed in real time. For example, if an altitude adjustment is suggested, the previously rendered safe altitude plane will be raised accordingly; if a modification to the flight path shape is suggested to bypass conflict areas, the recommended flight path skeleton will actively deform to generate a new path that avoids the heat map. This series of dynamic changes in visual elements constitutes the reconstructed intent guidance information. Finally, an interactive closed-loop adjustment phase begins. Users can directly accept the recommended reconstructed intent guidance information, or, with the assistance of the heat map, further refine the flight path range, shape, or key points through hand-drawn operations. Any user modifications will be captured and immediately trigger a new round of background calculations. This process includes recalculating the dynamic semantic potential field, as new flight path intentions may change the semantic weights of ground features; then, new candidate flight path segments are generated and submitted again to the collaborative management center for virtual spatiotemporal resource pre-occupancy and conflict prediction. This cycle will continue until the conflict prediction result returned by the collaborative management center shows that the spatiotemporal overlap is below the preset safety threshold. Only then will the conflict be considered resolved, and the user will be awaited to issue the final confirmation command.

[0045] For example, assuming the spatiotemporal overlap calculated in the previous step exceeds the safety threshold, a conflict prediction alarm is returned. A visualized spatiotemporal conflict heatmap is then overlaid and rendered on the electronic map, accurately marking the conflict area, and a dynamic adjustment suggestion message pops up: "It is recommended to delay the mission start time by 2 minutes." With the assistance of the heatmap, law enforcement officers did not adopt the delay suggestion. Instead, they dragged the entire orbital trajectory in the intention guidance information to raise the flight altitude from 65 meters to 80 meters. This user modification immediately triggered a new round of dynamic semantic potential energy field calculation and candidate flight path fragment generation. The new flight path was submitted to the collaborative management center for verification, and the result showed that the spatiotemporal overlap was 0, below the safety threshold, and the conflict was successfully resolved. The user was prompted that the flight path was safe and awaiting final confirmation. Through the spatiotemporal conflict heatmap and dynamic adjustment suggestion, the complex four-dimensional spatiotemporal conflict problem is presented to the user in an extremely intuitive way, allowing the user to instantly understand the problem and solution, demonstrating a high degree of intelligence and user-friendliness.

[0046] The task encapsulation module is used to optimize and encapsulate the candidate route segments that have been determined through conflict prediction after receiving a user confirmation instruction, generate the final route task instruction, and issue it for execution.

[0047] Optionally, the task encapsulation module includes: The conflict determination unit is used to receive the conflict prediction result of the candidate route segment returned by the collaborative management center. If the conflict prediction result is passed, the optimization process is triggered. The route fine-tuning unit is used to fine-tune and optimize the waypoint order and flight parameters in the candidate route segments based on the optimization process, and generate the optimized final route. The protocol encapsulation unit is used to encapsulate the waypoint coordinates and mission instructions of the optimized final route according to the protocol format that the target UAV can parse, and generate the final route mission instructions.

[0048] Specifically, this process is initiated by the conflict determination unit upon receiving the user's final confirmation instruction. This unit continuously monitors feedback from the collaborative management center. When it receives a conflict prediction result for the current candidate route segment and confirms the result as "passed" (i.e., the spatiotemporal overlap is below the safety threshold), the unit is activated. Once activated, it unlocks the route segment and triggers the subsequent route fine-tuning unit, initiating the final optimization process. If the received result is "conflict" or no feedback is received for an extended period, the unit remains locked and prompts the user to either re-plan the route or wait. Next, the route fine-tuning unit is activated. This unit does not re-plan but performs local optimization within the existing route framework. Fine-tuning operations include adjusting waypoint order and flight parameters. For example, for a coverage inspection task, the unit may reorder the waypoint sequence based on the importance distribution of ground features to ensure priority flight over critical areas; simultaneously, it will check and merge waypoints that are too densely packed or redundant on smooth tracks to reduce unnecessary acceleration, deceleration, and turning maneuvers. Fine-tuning of flight parameters focuses on energy consumption and safety. For example, cruise speed is appropriately increased in tailwind segments, while speed is reduced and altitude is increased in segments approaching obstacles. If the base cruise speed is 5 m / s, the real-time wind speed is 3 m / s, and the angle between the heading and wind direction is less than 30°, the speed is increased to approximately 6.2 m / s according to the aforementioned speed adjustment formula to utilize wind power to reduce energy consumption and shorten flight time. Conversely, in segments approaching obstacles, the speed is reduced to below 2.5 m / s based on the obstacle influence radius identified by the dynamic semantic potential energy field and safety margin requirements. At the same time, based on the obstacle height and the upper limit of UAV performance, the flight altitude is increased to 15-20 meters above the top of the obstacle to ensure a smooth transition of the flight path and maintain a sufficient safety margin. After this series of fine-tuning, the optimized final flight path is generated. Finally, the protocol encapsulation unit performs the final tasks of task packaging and distribution. This unit first serializes and encodes each waypoint in the optimized final flight path, including its longitude, latitude, and altitude coordinates, as well as commands such as flight speed, hovering time, and gimbal actions between waypoints, according to the target UAV's communication protocol, such as the format defined by the MAVLink micro-aircraft communication link protocol. Then, it encapsulates this encoded data into a complete mission file or data packet. This data packet not only contains flight path information but may also include metadata such as mission ID, takeoff time, and emergency return-to-home point coordinates. After encapsulation, the final flight path mission command is generated and transmitted to the designated UAV platform via wireless data link, awaiting UAV execution.

[0049] For example, after the user confirms a conflict-free flight path, the task encapsulation module is activated. The conflict determination unit confirms the "pass" result returned by the collaborative management center, triggering the optimization process. The flight path fine-tuning unit makes refined adjustments to the candidate flight path, for example, merging three overly dense redundant waypoints on the shaded side of the flight path, and fine-tuning the speed from 2.98 m / s to 3.2 m / s in the tailwind segment to optimize energy consumption. Finally, the protocol encapsulation unit serializes and encodes the three-dimensional coordinates, flight speed, gimbal pitch angle, and other commands of each waypoint in the optimized final flight path according to the MAVLink 2.0 protocol format supported by the target UAV, and encapsulates them together with metadata such as the task ID into a standard .plan task file. This final flight path task command is sent to the designated UAV via wireless data link, completing a seamless connection from planning to execution. By merging redundant waypoints and dynamically adjusting flight parameters, further optimization of flight efficiency and energy consumption is achieved, establishing a complete link from the upper-level intelligent planning system to the lower-level UAV flight control, realizing an automated closed loop for the entire process.

[0050] Optionally, the fine-tuning and optimization of the waypoint order and flight parameters in the candidate route segment includes: Adjust the waypoint density based on the dynamic semantic potential energy field, merge redundant waypoints and enhance the coverage of key areas to form a semantically coherent waypoint sequence. Based on the semantically coherent waypoint sequence, and combined with the environmental state information, the flight altitude and speed parameters between each waypoint are adjusted to achieve smooth flight path and energy consumption optimization. Based on the fine-tuned flight parameters and multi-aircraft collaborative constraints, the execution sequence of waypoints is optimized in time and aligned with the airspace to generate the optimized final route.

[0051] Specifically, the waypoint density and sequence are first adjusted based on a dynamic semantic potential energy field. This involves traversing all waypoints on candidate flight path segments and evaluating the semantic potential energy value of each waypoint's location and its surrounding area. For areas with gentle and low potential energy changes, such as open, safe airspace, a waypoint thinning strategy is employed. This involves calculating the vertical distance from a waypoint to the line connecting it to its preceding and following waypoints. If this distance is less than a preset redundancy threshold, such as 1 meter, the intermediate waypoint is merged to form a longer straight flight segment. Conversely, in critical areas with drastic potential energy gradient changes or extremely low potential energy values, such as mission target points or the edges of complex obstacles, waypoints are automatically densified. Cubic spline interpolation is used to increase intermediate waypoints, enhancing the accuracy and fit of the flight path to critical areas. This ultimately results in a more spatially distributed and logically coherent semantically coherent waypoint sequence. Based on this, and combined with real-time environmental state information, the flight parameters between waypoints are optimized. It analyzes each segment of the waypoint sequence and, combined with real-time environmental information such as wind speed and direction, calls a pre-set UAV energy consumption model to calculate a more suitable flight speed and altitude for each segment. For calculating the new flight speed of segment i... and the new flight altitude of flight segment i ,have: ; in, Minimum safe flight speed; The maximum permissible flight speed is subject to limitations imposed by the drone's performance and air traffic control. This is the nominal flight speed, the economic cruising speed under windless conditions; This is the semantic potential influence factor for flight segment i, typically ranging from 0.7 to 1.3. Base cruising altitude; To provide a safety margin based on the type of ground features and potential energy values ​​below; The wind field height correction factor is typically set to 5-15 meters. Finally, based on the fine-tuned flight parameters and multi-aircraft coordination constraints obtained from the coordination management center, the execution sequence of waypoints is optimized in terms of timing and airspace alignment. This process uses the multi-aircraft coordination scheduling table issued by the coordination management center as a benchmark, combined with the expected execution time window and airspace occupancy profile of the local flight segment, to dynamically rearrange and timestamp the waypoint sequence. Specifically, based on task priority and airspace sharing agreement, overlapping flight segments are adjusted in the time dimension to avoid peak times, such as inserting controllable delays into non-critical flight segments or implementing altitude stratification for parallel flight segments; in the spatial dimension, the three-dimensional coordinates of waypoints are slightly shifted according to airspace alignment rules to ensure that vertical or horizontal spacing with the cooperating aircraft group is maintained within a safe margin in the shared airspace. The optimized waypoint sequence will be accompanied by a trigger timestamp accurate to milliseconds and an airspace alignment identifier, generating an optimized final flight route that is consistent with the multi-aircraft system coordination in terms of time, space, and behavioral parameters. Through this precise temporal arrangement and proactive spatial alignment, a final route is generated that is optimized in terms of semantics, energy consumption, dynamics, and spatiotemporal coordination.

[0052] For example, before final encapsulation, the route fine-tuning unit performs refined optimization of candidate route segments. First, it adjusts the waypoint density based on a dynamic semantic potential energy field. In the "core reconnaissance target" area near illegal buildings, waypoints are added through interpolation to enhance coverage; in open turning areas, redundant waypoints are merged to smooth the trajectory. Next, combining real-time wind speed and other environmental information, it calls the energy consumption model to fine-tune the flight speed and altitude of each segment, such as appropriately reducing speed in headwinds, thereby optimizing energy consumption. Finally, it performs temporal optimization and airspace alignment. To avoid potential conflicts with another delivery drone that is about to pass, a 10-second hovering waiting instruction is proactively added at the last waypoint before reaching the public passage. Through this series of fine-tuning at the semantic, energy consumption, dynamic, and spatiotemporal coordination levels, a highly optimized final route is generated. The dynamic adjustment of flight parameters by combining environmental information and the energy consumption model improves energy efficiency.

[0053] Based on the same inventive concept, such as Figure 5 As shown, the present invention also provides a method for generating flight paths and adapting to multiple scenarios for government drones based on hand-drawn diagrams, the method comprising: The system acquires the user's handwriting data on the electronic map, the current drawing area, and the selected government task type data. It then performs real-time intent analysis on the handwriting data and the government task type data to generate intent guidance information, which is then displayed on the electronic map. In response to the intent guidance information, the current drawing area is subjected to feature recognition and semantic enhancement perception, and the identified features are dynamically assigned semantic attributes based on the government task type data. A dynamic semantic potential energy field is generated based on the semantic attributes of all features. Based on the dynamic semantic potential energy field, real-time route simulation is performed to generate candidate route segments, and the candidate route segments are submitted to the collaborative management center for virtual spatiotemporal resource pre-occupation and conflict prediction. Upon receiving a user confirmation instruction, the candidate route segments identified through conflict prediction are optimized and encapsulated to generate the final route task instruction, which is then issued for execution.

[0054] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0055] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A system for generating flight paths and adapting to multiple scenarios for government drones based on hand-drawn diagrams, characterized in that, The system includes: The intent parsing module is used to obtain the user's handwriting data, the current drawing area, and the selected government task type data on the electronic map, perform real-time intent parsing on the handwriting data and the government task type data, generate intent guidance information, and display it on the electronic map. The semantic perception module is used to respond to the intent guidance information, perform feature recognition and semantic enhancement perception on the currently drawn area, and dynamically assign semantic attributes to the identified features in combination with the government task type data, and generate a dynamic semantic potential energy field based on the semantic attributes of all features. The route pre-simulation module is used to perform real-time route simulation based on the dynamic semantic potential energy field, generate candidate route segments, and submit the candidate route segments to the collaborative management center for virtual spatiotemporal resource pre-occupation and conflict prediction. The task encapsulation module is used to optimize and encapsulate the candidate route segments that have been determined through conflict prediction after receiving a user confirmation instruction, generate the final route task instruction, and issue it for execution.

2. The system for generating flight routes and adapting to multiple scenarios for government drones based on hand-drawing, as described in claim 1, is characterized in that... The intent parsing module includes: The intent inference unit is used to analyze the morphological features of the handwriting data and, in combination with the government task type data, infer the user's potential work intent. The rule loading unit is used to load associated flight rules and operational constraints from a preset scenario strategy knowledge base according to the potential operational intent. The guidance generation unit is used to generate and render intent guidance information in real time based on the flight rules and operational constraints, as well as the identified key ground features related to the current operational intent.

3. The system for generating flight routes and adapting to multiple scenarios for government drones based on hand-drawing, as described in claim 2, is characterized in that... The loading of associated flight rules and operational constraints from the preset scenario strategy knowledge base includes: Based on the potential task intent and the geographical features of the currently drawn area, multi-level semantic retrieval and context matching are performed in the scene strategy knowledge base to dynamically construct a task scene context graph; Based on the operational scenario context graph, flight behavior templates that match the administrative task type are extracted; The system acquires real-time environmental status information and performs online corrections and flexible scaling of behavior parameters on the flight behavior template to form real-time effective flight rules and operational constraints.

4. The system for generating flight routes and adapting to multiple scenarios for government drones based on hand-drawing, as described in claim 1, is characterized in that... The semantic awareness module includes: The topology construction unit is used to identify the basic geographic features within the currently drawn area, and combine the intent guidance information to perform enhanced perception and structured analysis to construct a topology network that reflects the relationship logic between geographic features; The semantic matching unit is used to dynamically match semantic attributes of the various feature nodes identified in the topological network with the task scenario based on the government task type data, forming a semantic feature set; The potential energy field generation unit is used to calculate the semantic force field distribution and generate a dynamic semantic potential energy field based on the semantic role association degree of each element in the semanticized feature set and its influence weight on task execution.

5. The system for generating flight routes and adapting to multiple scenarios for government drones based on hand-drawing, as described in claim 1, is characterized in that... The route pre-simulation module includes: The path exploration unit is used to explore paths based on the potential energy gradient in the dynamic semantic potential energy field and generate an initial path point sequence. The route calibration unit is used to perform smoothing optimization and feasibility verification based on the initial path point sequence to form a basic route segment. The evaluation and submission unit is used to calculate the performance indicators of the basic route segment, assign corresponding priority weights to the performance indicators according to the government task type data, obtain weighted evaluation results for configuration selection, generate candidate route segments, and submit them to the collaborative management center for virtual spatiotemporal resource pre-occupancy and conflict prediction.

6. The system for generating flight routes and adapting to multiple scenarios for government drones based on hand-drawing, as described in claim 5, is characterized in that... The process of submitting virtual spatiotemporal resource pre-occupation and conflict prediction to the collaborative management center includes: Extract the spatiotemporal resource occupancy information contained in the candidate route segments, the spatiotemporal resource occupancy information including the planned occupancy time window and the three-dimensional contour of the flight airspace; The spatiotemporal resource occupancy information is sent to the collaborative management center, and the spatiotemporal overlap is calculated with the spatiotemporal resource occupancy information of other tasks to obtain the spatiotemporal overlap. If the spatiotemporal overlap exceeds a preset safety threshold, a conflict prediction alarm is generated and returned as feedback information to adjust the intent guidance information.

7. The system for generating flight routes and adapting to multiple scenarios for government drones based on hand-drawing, as described in claim 6, is characterized in that... The adjustment of the intent guidance information includes: Based on the spatiotemporal overlap area, conflict task type, and current task scenario contained in the conflict prediction and alarm information, a visual spatiotemporal conflict heat map and dynamic adjustment suggestion information are generated. The visualized spatiotemporal conflict heatmap is overlaid and rendered onto the currently drawn area of ​​the electronic map, and the intent guidance information is reconstructed in real time based on the dynamic adjustment suggestion information to obtain the reconstructed intent guidance information. In response to the user's secondary confirmation or modification of the reconstructed intent guidance information, a new round of dynamic semantic potential energy field calculation and candidate route fragment generation is triggered until the spatiotemporal overlap is lower than the safety threshold.

8. The system for generating flight routes and adapting to multiple scenarios for government drones based on hand-drawing, as described in claim 3, is characterized in that... The task encapsulation module includes: The conflict determination unit is used to receive the conflict prediction result of the candidate route segment returned by the collaborative management center. If the conflict prediction result is passed, the optimization process is triggered. The route fine-tuning unit is used to fine-tune and optimize the waypoint order and flight parameters in the candidate route segments based on the optimization process, and generate the optimized final route. The protocol encapsulation unit is used to encapsulate the waypoint coordinates and mission instructions of the optimized final route according to the protocol format that the target UAV can parse, and generate the final route mission instructions.

9. A system for generating flight routes and adapting to multiple scenarios for government drones based on hand-drawing, as described in claim 8, is characterized in that... The fine-tuning and optimization of the waypoint order and flight parameters in the candidate route segments includes: Adjust the waypoint density based on the dynamic semantic potential energy field, merge redundant waypoints and enhance the coverage of key areas to form a semantically coherent waypoint sequence. Based on the semantically coherent waypoint sequence, and combined with the environmental state information, the flight altitude and speed parameters between each waypoint are adjusted to achieve smooth flight path and energy consumption optimization. Based on the fine-tuned flight parameters and multi-aircraft collaborative constraints, the execution sequence of waypoints is optimized in time and aligned with the airspace to generate the optimized final route.

10. A method for generating and adapting government drone flight paths based on hand-drawing, and adapting to multiple scenarios and tasks, comprising the system for generating and adapting government drone flight paths based on hand-drawing as described in any one of claims 1-9, characterized in that... The method includes: The system acquires the user's handwriting data on the electronic map, the current drawing area, and the selected government task type data. It then performs real-time intent analysis on the handwriting data and the government task type data to generate intent guidance information, which is then displayed on the electronic map. In response to the intent guidance information, the current drawing area is subjected to feature recognition and semantic enhancement perception, and the identified features are dynamically assigned semantic attributes based on the government task type data. A dynamic semantic potential energy field is generated based on the semantic attributes of all features. Based on the dynamic semantic potential energy field, real-time route simulation is performed to generate candidate route segments, and the candidate route segments are submitted to the collaborative management center for virtual spatiotemporal resource pre-occupation and conflict prediction. Upon receiving a user confirmation instruction, the candidate route segments identified through conflict prediction are optimized and encapsulated to generate the final route task instruction, which is then issued for execution.

Images